The project addresses two key challenges in selfie skin analysis: creating consistent images for accurate AI analysis and protecting users' identities without losing relevant skin details.
Dermaself's AI skin analysis technology transforms a selfie into an analysis of visible skin characteristics and uses the results to generate personalised skincare recommendations from a brand's or retailer's product catalogue.
But using selfies for skin analysis presents two major technical challenges: image consistency and privacy.
Through a direct collaboration between academia and industry, five Alta Scuola Politecnica students developed an image-processing infrastructure designed to tackle both.
Why selfie quality matters for AI skin analysis
A smartphone selfie may look perfectly normal to the human eye, but cameras do not always represent skin in the same way.
Automatic exposure, white balance, camera processing and environmental lighting can change the appearance of skin from one image to another. These variations can affect the quality and consistency of AI-powered skin analysis, particularly when algorithms need to identify subtle visible characteristics such as spots, redness, lesions or skin texture.
For an AI skin analysis platform, standardising images before analysis can therefore help create more consistent and comparable inputs.
The challenge becomes even more important when skin analysis is used across different environments. A consumer might take a selfie at home, while another uses the same technology in a beauty store, during an event or through an e-commerce experience.
The underlying AI needs to work with images captured under all these different conditions.
The privacy challenge behind selfie skin analysis
There is also another fundamental issue: privacy.
A facial image contains two different types of information at the same time. It shows the characteristics of a person's skin, but it can also reveal their identity.
Traditional facial anonymisation techniques, such as blurring, masking or pixelating a face, can protect identity. However, they can also remove precisely the skin details required for AI skin analysis and dermatological research.
This creates an important challenge for beauty tech and digital dermatology: how can a facial image be anonymised without destroying the skin information that makes it useful?
The Alta Scuola Politecnica team developed a solution specifically designed to address this problem.
One research team across Politecnico di Milano and Politecnico di Torino
The project brought together five master's students from the two universities behind Alta Scuola Politecnica.
Andrea Germano and Adriano Giuliani, studying Computer Engineering, and Federico Greppi, studying Biomedical Engineering, represented Politecnico di Torino.
Edoardo Gribaldo, studying Computer Science and Engineering, and Alessia Soccionovo, studying Management Engineering, represented Politecnico di Milano.
The project was guided by academic tutors Professor Elisabetta Raguseo of Politecnico di Torino, who is also Director of Alta Scuola Politecnica, and Professor Federica Arrigoni of Politecnico di Milano.
Working directly with Dermaself, the team focused on developing technology that could move beyond academic research and become part of a real-world AI skin analysis platform.
Standardising selfies before AI skin analysis
The first part of the solution focuses on selfie image normalisation.
Before the photo is captured, the system helps guide the user towards suitable conditions, including the correct distance and lighting.
After capture, the technology corrects remaining variations in illumination through a hybrid image-processing pipeline combining AI-driven illumination correction with classical algorithms.
The objective is straightforward: provide Dermaself's AI with more consistent images, regardless of where or how the selfie was taken.
By reducing variations caused by lighting and camera conditions, the technology creates more standardised inputs for the algorithms used to analyse visible skin characteristics.
Anonymising a face without losing skin detail
The second part of the project addresses one of the most complex questions surrounding selfie skin analysis and privacy: preserving useful skin information while removing identifiable facial characteristics.
Instead of simply blurring or masking the face, the team developed a 3D-based facial reconstruction and anonymisation pipeline.
The system reconstructs the user's face in three dimensions and modifies its underlying geometry to reduce biometric identifiers. The original high-resolution skin texture is then re-projected onto the modified facial structure.
This approach is designed to preserve visible skin information such as spots, lesions, redness and texture, while making the person represented in the image no longer recognisable.
In practical terms, the technology aims to separate two elements that are normally inseparable in a selfie: identity and skin information.
From academic research to Dermaself's AI skin analysis technology
The technologies developed through the Alta Scuola Politecnica project are designed to work within Dermaself's AI skin analysis ecosystem.
Dermaself enables beauty brands and retailers to integrate selfie-based skin analysis into digital and physical customer experiences. From a selfie, the technology analyses visible skin characteristics and connects the results with products from the company's own catalogue to create personalised skincare recommendations.
The technology can be integrated into different touchpoints, including e-commerce platforms, beauty retail environments and brand events.
Image standardisation can help the system process selfies more consistently across these environments, while anonymisation provides a new approach to protecting the identity of the people represented in those images.
Early results from the research are promising, with broader validation, including evaluation by dermatologists, representing the next stage of the project.
Why privacy-preserving AI skin analysis matters for beauty tech
Privacy-preserving image technology could have applications beyond individual skincare recommendations.
Beauty companies, research organisations and clinics increasingly need high-quality image datasets to develop and validate AI skin analysis and digital dermatology technologies.
However, collecting and sharing facial images creates significant privacy challenges.
A system capable of preserving relevant skin characteristics while reducing identifiable biometric information could make it easier to create, store and potentially license datasets for research and AI development while better protecting the individuals represented in them.
The project therefore addresses a broader question facing the beauty industry: can AI skin analysis become more accurate without compromising consumer privacy?
The work developed by Alta Scuola Politecnica with Dermaself demonstrates one possible technical approach, combining image standardisation, 3D facial reconstruction and privacy-preserving image processing.
Above all, the collaboration shows how academic research and beauty tech companies can work together to solve real industry challenges, translating university research into technology designed for real-world applications.
Frequently Asked Questions
What is AI skin analysis?
AI skin analysis uses artificial intelligence and computer vision to analyse visible skin characteristics from an image, such as a smartphone selfie. Depending on the technology, the results can be used to identify visible skin concerns and support personalised skincare recommendations.
How does Dermaself's AI skin analysis work?
Dermaself analyses visible skin characteristics from a selfie and uses the results to recommend skincare products from a brand's or retailer's own catalogue. Product recommendations are generated by connecting skin analysis data with product and ingredient information.
Why does selfie lighting matter for AI skin analysis?
Lighting, exposure and white balance can change how skin appears in a photograph. Standardising these variables helps provide more consistent images for AI skin analysis and makes selfies captured in different environments more comparable.
How can a selfie be anonymised without losing skin information?
The technology developed by the Alta Scuola Politecnica team uses a 3D-based facial reconstruction pipeline. Instead of blurring the image, it modifies the underlying facial geometry while re-projecting the original high-resolution skin texture. This is designed to reduce identifiable facial characteristics while preserving visible skin details needed for analysis.
Who developed the selfie normalisation and anonymisation technology?
The technology was developed by a multidisciplinary team of five master's students from Alta Scuola Politecnica, the joint honours programme of Politecnico di Milano and Politecnico di Torino, in collaboration with Dermaself and under the guidance of academic tutors from both universities.
What is selfie image normalisation?
Selfie image normalisation is the process of reducing differences caused by factors such as lighting, exposure and camera processing. For AI skin analysis, this can help create more consistent inputs and improve the comparability of images captured under different conditions.
Why is privacy important in AI skin analysis?
A facial image can contain both information about a person's skin and information that can identify the individual. Privacy-preserving technologies aim to separate these two elements so that useful skin information can be retained while reducing exposure of identifiable facial characteristics.
Can this technology be used to create datasets for AI research?
Privacy-preserving facial images could potentially support the creation, storage and sharing of datasets for beauty tech and digital dermatology research. Broader validation of the technology is still required, including review by dermatologists.
